IV. ParMRC and Expandable DNA Nanostructures
IV.2. Results and Discussion
IV.2.2. Expandable nanostructures
IV.2.2.1. Model system: origami as ParR/parC cores stabilizing ParM filaments 101
101
102
microscope, the fluorescent dyes (Alexa488) need to be excited by light source of proper wavelength (~488 nm). Upon light irradiation, while we could see a number of filaments that had been grown in that area, we observed the filaments suddenly shrinking, breaking and disappearing out of the field (Figure IV-13). When moved to a nearby area, there were again a number of filaments, which again soon disappeared. We knew that the filaments had grown, but we just could not capture the moment of growing, because as soon as we shined light, the grown filaments broke and disappeared. It seemed to be photo-damaging rather than pure photo-bleaching, because instead of the whole field uniformly darkening, the filaments tended to break in the middle and shrunk from their ends, with small pieces quickly diffusing away. To overcome the problem, reducing the light intensity by using neutral density filters was essential, but it sacrificed the signal as well—there was a tradeoff between signal and photo-damage. A less destructive imaging system, total internal reflection fluorescence (TIRF) microscopy, which uses a narrow window (~200 nm) of evanescent wave rather than direct light irradiation, hence exciting molecules only near the surface and minimizing disturbance to the remaining system, was also tried. However, now that the depth of view was limited near the surface, the surface background became more of an issue. Even with a narrower window of excitation, the signal-to-noise ratio did not improve, perhaps because surface- bound fluorescent monomers provided high background. Proper surface treatment that can discriminate filaments above a certain length from monomers needed to be devised.Figure IV-13. ParM filaments that break, shrink and disappear from the field of imaging by photo- damaging. Time points are indicated at the top of each frame. Note the rapid time scale. The longest filament (~5 um) shown in this field dissociates completely within ~0.4 s, which corresponds to a koff of ~5×103 /s, about 100-fold higher than normal.
Nonetheless, despite those limitiyng conditions, we were able to capture the growing moment of some filaments. It was still a very rare event, which required analysis of multiple movies of filaments, the majority of which were merely shrinking. The signal-to-noise ratio was still low.
One representative filament that grew for a while then shrunk is shown in Figure IV-14. From the
5 um
t=0s 0.232s 0.290s 0.405s
103
time lapse images, we could obtain the length change curves for the growth and shrinkage phases for this particular filament (Figure IV-14b). From linear fits to the curves, we estimated the rate constants: kon = ~9.9 /uM/s, and koff = ~27.7 /s. These values are within the same order as the known values (5 /uM/s and 64 /s, respectively97), but the on rate is higher and the off rate is lower, each by a factor of ~2. This might suggest that the surface is giving extra stabilization to ParM monomers, which is consistent with the high background observed.Figure IV-14. Dynamic instability of a ParM filament. (a) Representative frames from a movie that captured the growth and shrinkage of a ParM filament. (b) Plot showing the length change and the linear fits for the growth phase and the shrinkage phase.
IV.2.2.1.2. Colocalization of ParM filaments with DNA origami containing ParR/parC complexes
After we confirmed that ParM behaves as expected, exhibiting dynamic instability, and we could observe the behavior under our instrument (with some limitations), we moved forward to mixing ParM filaments and DNA origami that contain parC strands, along with ParR and ATP. We
t=12.7s
1 um
35.0s 30.6s
23.6s
21.7s 17.2s
Length (um)
Time (sec)
Growth Shrinkage
(a)
(b)
y = -0.0678x + 2.8273 R² = 0.84211
20 25 30 35 40
y = 0.0558x - 0.1519 R² = 0.8855
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0 5 10 15 20 25
104
labeled DNA origami with Cy3 and ParM with Alexa488. Although we still struggled capturing dynamic behavior of ParM filaments, we were able to observe some colocalization of DNA origami containing ParR/parC complexes and the ends of ParM filaments in static images (Figure IV-15), suggesting that the ParR/parC complexes do engage with the ends of ParM filaments.Figure IV-15. Colocalization of the ends of ParM filaments and DNA origami containing ParR/parC complexes. Majority of the origami with ParR/parC complexes (labeled red) appear at the ends of ParM filaments (labeled green), suggesting the binding of ParM filaments to ParR/parC complexes. Scale bars: 5 um.
IV.2.2.1.3. Reference system with microbeads in place of origami
While we could get some evidence that individual components work properly—ParR binding to parC sites on origami with specificity, and ParM filaments dynamically growing and shrinking—
and we could witness colocalization of origami containing ParR/parC complexes with the ends of ParM filaments, as discussed in several previous sections, we could not successfully get any dynamic behavior of the system when they were mixed altogether. Could it be the DNA origami structure, e.g., the size, aspect ratio, etc., that prevents the system from working? To test the hypothesis that the DNA origami might cause some problem and not other components, we decided to take a step back and reconfirm that all other components worked fine by using micron-sized beads in the place of origami, which is closer to what Garner et al. tested in their first reconstruction system105. We used fluorescent beads (Nile Red—whose emission spectrum is wide and lies across both the Alexa488 channel and Cy3 channel) with ~1 um diameter, coated with streptavidin. We had biotin at the end of each parC strand (see Materials and Methods), so we were able to coat the beads with the parC strands via the streptavidin-biotin interaction.
Interestingly, the use of microbeads revealed the most important problem—surface treatment. Even with the beads, our initial attempts to get dynamic behavior of ParM filaments failed. Meanwhile, we often observed that the beads were stuck to a glass surface, and even
105
witnessed the cases where beads and filaments were stuck on a different surface (either the top—coverslip—or the bottom—glass slide). This led us to the realization that the surface might have been the main issue—not only creating high background signal (which was noted earlier), but more importantly, being too sticky to the molecules that it distorted the concentrations of all the components. We had been following the standard protocol for surface passivation suggested by the group that provided the proteins, which is to add bovine serum albumin (BSA) into the system such that it dominates the surface and blocks other proteins from binding. But that protocol did not seem to be optimal at least for our system. Alternatively, we tried to use the older protocol that was originally used in the 2007 Garner et al. reconstruction work, which is chemically treating the surface to make it highly hydrophobic (silanization using diethyl-dichloro-silane; see Materials and Methods). Using the chemically treated glass slides and coverslips, we could finally observe a dynamic behavior of ParM filaments, pushing apart microbeads coated with ParR/parC. Some frames are captured in Figure IV-16, and the full movie is available for download at http://dna.caltech.edu/Woo-thesis-movies.
Figure IV-16. Frames from a movie that shows networks of ParM filaments/bundles growing and pushing microbeads apart. Time points are indicated at the top of each frame. The beads show up as dots that are brighter than the filamentous parts.
IV.2.2.1.4. Large scale expanding network structures in the high origami concentration regime
Now, with the surface problem solved, we went back to testing the system with DNA origami. Using triangular origami at ~400 pM, each containing 177 parC strands, we were able to capture dynamically expanding ParM filaments (Figure IV-17, the full movie is available at http://dna.caltech.edu/Woo-thesis-movies). Although each origami core is not identifiable because
10 um
t=0s 36s 91s 164s
106
they were not fluorescently labeled, the fact that the expanding filaments were stabilized by origami is supported by a control experiment where, in the absence of origami, ParM filaments that potentially form are not observable under the same setup, perhaps due to their transient nature and narrow thickness.Figure IV-17. Frames from a movie that shows networks of ParM filaments/bundles which grow and expand in the presence of DNA origami with ParR/parC complexes. Time points are indicated at the top of each frame. DNA origami were not fluorescently labeled, so they do not appear.
IV.2.2.1.5. Distinct filament bundle stabilization in the low origami concentration regime: Exploring the vast parameter space, especially the underestimated role of the crowding agent
Although the expansion of ParM filament bundles stabilized by DNA origami ParR/parC cores described in the previous section was promising, the system is not ready to be used for expandable structures because the filament bundles tended to form large scale network structures with one another. We wanted to find a regime where filaments/bundles do not form extensive networks and are stabilized by one origami containing clusters of ParR/parC complexes at each end.
Finding the right regime required exploring a large, multi-dimensional parameter space of concentrations of all different components—much larger than we initially anticipated. We initially thought that the space would be defined by the concentrations of just the three essential components—parC, ParR, and ParM. In fact, even within just the three-dimensional space, there were a large number of cases to try, to find a regime that works reliably and in a somewhat expected way. Yet, there were more variables that we had not considered important.
The most influential variable was the concentration of the crowding agent, methylcellulose in our system. Methylcellulose is often used for microscopy samples that contain molecules with tubular shapes to help confine them near slide/coverslip surfaces106, while not letting them permanently stick to the surfaces. Methylcellulose can also cause molecules with tubular shape to
10 um
t=0s 37s 111s 130s
107
bundle together107 by volume exclusion. The effect of methylcellulose on bundle formation also applies to ParM filaments; the higher the concentration of methylcellulose, the more bundle formation is promoted. But we had not realized that the concentration of methylcellulose was a critical parameter for even “observing” the filament bundles under the fluorescence microscope. We later learned that without using methylcellulose, ParM filament bundles do not get stabilized and the in vitro reconstructed ParMRC system does not work108. A small increase of methylcellulose (by~0.4% of total solution volume) induced too much stabilization to the bundles, causing spontaneous bundle formation and stabilization of the filaments even without ParR/parC complexes at the ends.
In addition, the fraction of fluorescent ParM among the total ParM monomer population was another important variable. Adding fluorescently labeled ParM is essential for the ability to observe their behaviors, and the fraction of fluorescent ParM needed to be optimized to obtain the best signal-to-noise ratio and a signal level comparable to that from the other channel for origami or beads. However, the fluorescent tags may somehow affect the behavior of the individual protein, so using high fractions of them may change the overall behavior of the system. This turned out to be true in our experiments; with 100% labeled ParM, filaments seemed to become overly stabilized even without any other stabilizing factors (e.g., ParR/parC or AMPPNP, a non-hydrolyzable ATP analog), perhaps suggesting the fluorescence labeling might have disabled the ATPase activity of some portion of the ParM monomers, or otherwise changed the monomers to favor the configuration in filaments and bundles.
After exploration of the large parameter space, we found a regime where distinct filament bundles are stabilized by ParR/parC complexes on origami at each end. To get a smaller number of
“stabilization cores”, a lower concentration of DNA origami was used (~50 pM, with total [parC] =
~10 nM). Also, the concentration of methylcellulose was reduced (from 0.8% to 0.4 %) to suppress spontaneous bundle stabilization. The concentration of ParM monomer was increased (~7 uM) to increase the thickness of bundles that do get specifically stabilized by ParR/parC and hence improve the signal-to-noise ratio. Some frames from a representative movie are shown in Figure IV-18, and the full movie is available at http://dna.caltech.edu/Woo-thesis-movies.
108
Figure IV-18. Frames from a movie that shows distinct ParM filaments/bundles whose ends are stabilized by DNA origami with ParR/parC complexes. Time points are indicated at the top of each frame.
At the regime of lower concentration of origami and higher concentration of ParM monomers (than in the case of Figure IV-17), more filaments are anchored at each origami, forming thick bundles of filaments. In this series of images, two independent filament bundles appear, whose ends are visible. The ends of the top bundle are indicated by orange arrowheads and the ends of the bottom bundle are indicated by blue arrowheads. The two ends meet and the top bundle continues to grow along the bottom bundle. The bottom bundle also grows, which can be observed by the relative movement of the bottom end of the bundle.
IV.2.2.2. Expanding origami chains
Now that we found the working regime for the ParMRC system mediated by DNA origami, we moved on to testing expandable structures—once connected and later dynamically expanded by actively growing protein filaments. We first tested chains of rectangular origami connected by stacking bonds, because (1) they are easy to create—something that we know works very well and we are very familiar with, (2) stacking bonds we believe are a good candidate for reconfigurable bonds because they may have lower activation energy to mechanical deformation such as shear force, and (3) it is easy to detect the change in structure under fluorescence microscope due to its
t=20s 55s 60s 120s 145s 165s 205s 230s
5 um
109
large size in length.IV.2.2.2.1. Design of origami chains
Figure IV-19. Design and microscope images of DNA origami rectangle chains. (a) Schematic diagram of the DNA origami rectangle chains used for the study. 96 parC strands (depicted as vertical stretches of DNA, with smaller number than actual, for simplicity) were anchored in the central half area of each rectangle. Each of the parC strands was fluorescently labeled with Cy3. (b-e) AFM images and (f) a fluorescence microscope image that show the rectangle chains. Densely packed parC strands seem to induce twist over the length of the chains, generating somewhat periodic kinks, folds, and breaks when deposited on a surface (indicated by arrows).
We designed staple strands with 5’-end extension (T20) to anchor parC strands for the rectangle origami. By choosing a subset of staple locations, we can control the total number and locations of the parC strands. For initial tests of expandable origami chains, we used 50% of the staples containing the anchor for parC strands (96 staples) at the center half area, far from the edges (Figure IV-19a). Each of the parC strands were fluorescently labeled with Cy3 (by hybridizing (CTT)5-Cy3 with the end part of the parC “bottom” strand). All rectangles were first made without any edge staples to prevent stacking, gel purified, then put through a second annealing step with edge staples to allow chain formation. Figure IV-19b-e show the AFM images of the origami chains where each origami rectangle contains parC strands in the central half area. While the rectangle
... ...
100 nm 200 nm 400 nm 2 um 2 um
(a)
(b) (c) (d) (e) (f)
110
used was a twist-corrected design (the same as that used in Chapter II), the introduction of parC strands seem to have induced some global twist or curvature into each rectangle and accumulated along the chains (noted by somewhat periodic kinks, folds or breaks in the chain structures, some of which are marked by arrows in Figure IV-19b-e), perhaps due to the repulsion between the densely packed, long parC strands. Figure IV-19f shows a fluorescence microscope image of the origami chain sample, whose field-of-view size is the same as the AFM image shown in Figure IV-19e.Although with lower resolution due to the diffraction limit, the chains in the fluorescence microscope image show similar structural features to those in the AFM images, e.g., kinks and rough length distribution.
IV.2.2.2.2. Can ParM filaments separate DNA origami? Estimation and comparison of force generated by ParM filament growth and force required to
separate origami in a chain
Now we ask: can growing ParM filaments separate the DNA origami rectangles in chains at all? Are the forces or energies comparable? In this section, we estimate the force generated by the growth of ParM filaments and the force required to break stacking bonds and separate origami in a chain, and compare them. The ambiguity in determining the length scale of stacking bond breakage makes it difficult to estimate the relevant force from the bond energy, and just comparing the chemical energies involved in the two systems—ParM polymerization and stacking bonds—might give a useful insight. Nonetheless, here we make attempts to estimate the forces based on two different assumptions about the relevant distance, one from a simulation-based stacking energy potential curve and the other by simply taking the distance of separation created by a ParM polymer at each insertion step. In fact, the second assumption makes the comparison effectively the same as comparing at the energy level.
Also, a more careful analysis than a mere force comparison would require additional considerations about the lifetimes of the bonds, i.e., the off rate of ParM monomers associated in filaments and the off rate of stacking bonds. For example, if a ParM monomer at a filament end dissociates at a rate much faster than the dissociation of stacking bonds, the filaments in a bundle might not be able to collectively exert enough force to separate stacking bonds. Conversely, if the off rate of stacking bonds is reasonably higher than the ParM monomer dissociation rate, the growth of ParM filaments will be able to successfully break the bonds within the lifetime of the associated monomers. We do not have a good estimate for the off rates of stacking bonds. (Although we have an estimate for the bond energy and hence the equilibrium constant, and a rough sense of the initial
111
rate of bond formation—they are observed to have formed stacking bonds within the order of minutes—we do not have an estimate for the on/off rates of the bond at its equilibrium state.) Also, the off rate of a ParM monomer from a filament end while it is capped by a ParR/parC complex is not known, although the off rate of an ATP-bound ParM (the state of a ParM monomer at a capped end) at free filament ends can be estimated from the Kd and the on rate of ATP-ParM: the Kd can be treated the same as the critical concentration of the monomers for polymerization systems109,110, and hence, using the critical concentration, 0.6 uM, and the on rate, ~5/uM/s, of ATP-ParM97, we can obtain the off rate of ATP-ParM at free ends to be ~3/s. However, once filament ends are stabilized by ParR/parC complexes, ParM filaments seem to grow continuously at least over the course of seconds to minutes, as shown in the continuous travel of ParR/parC complexes pushed by ParM polymerization from one pole to the other within a cell111 and in the steady growth of filament bundles in the in vitro reconstitution assay105, unless they break and shrink from the middle by buckling or photodamage. Hence, it seems reasonable to treat ParM filaments as a quasi-infinitely growing thermal “ratchet”109,110,112, as long as ParM monomers in solution maintain the steady state concentration, and thus, we limit our estimation and comparison to the amounts of relevant forces in our system.IV.2.2.2.2.1. Estimation of the force generated by ParM filament growth
The dynamics of a filament end, e.g. the on/off rates of a monomer, is likely to be affected by the presence of a ParR/parC cap, especially when the filament is exerting force109,110 to push ‘loads’, such as plasmids or DNA origami structures. For example, the monomer addition is likely to be inhibited by the presence of the cap and the load, decreasing the on rate, and the reaction force against pushing would likely destabilize the monomer at the end, increasing the off rate. The change, especially the increase, in Kd, can be considered to correspond to an extra free energy cost that arises from the work that the filament growth has to do against its load110. Hence, the dissociation constant in the presence of force, Kd (F) can be expressed by the following equation110:
Kd(F) = Kd(0)eΔΔG/kBT,
where Kd (0) is the native dissociation constant in the absence of force, ΔΔG is the extra free energy cost in the presence of force, kB is the Boltzmann constant, and T is the absolute temperature. Since the dissociation constant is the same as the critical concentration of monomers in polymerization systems, the Kd under force can be interpreted as the critical concentration required to balance the force (below the critical concentration, the filament would not grow). Hence, the maximum force, Fmax, that a filament end can ‘withstand’, or by reaction, ‘exert’ against a load over a distance δ at a monomer concentration M can be expressed by the following equation110: